Proceedings of the 11th Congress of International Society of Rock Mechanics, Lisbon, 9-13 July 2007, pp 979-982

ON THE CLASSIFICATION OF THE ROCK MASS EXCAVATION BEHAVIOUR IN TUNNELING

G. Russo & P. Grasso

Geodata SpA, Turin, Italy

ABSTRACT: This paper deals with the forecasting of excavation behaviour in tunneling and presents a new scheme for a
rational classification. In particular, a scheme is proposed essentially based on the combination of two classification
systems: the first is basically centred on the results of stress analysis, while the second, which is made up of the RMR
system, is specifically directed towards the representation of the geostructural characteristics of the rock mass and to the
relative self-supporting capacity. The matrix that results from this double-classification approach allows for first an optimal
focalisation of the design problem and then a rational choice of the stabilisation measures in function of the most probable
potential deformation phenomenon. A practical application is presented with probabilistic implementation of the method.
1 INTRODUCTION classification indexes that are able to express the potential
intensity of the expected deformation phenomena.
Almost ten years have passed since a joint initiative of the Reference to an “equivalent-continuous” geotechnical
main Italian associations working in the sector of tunneling, model is in such case generally implicit, whether making
developed the “Guidelines for the Design, Tendering and use of empirical or more analytical methods. In the former
case, a typical example is that of the “Competence
Construction of Underground Works” ("LGP", Indexes”, which express the relationship between a
NPSUW,1997). As can be deduced from the title itself, this representative term of the stress conditions at tunnel level
document did not intend to be a design manual, but rather a (stresses at the excavation boundary or the lithostatic
recommended procedure for the organic and complete pressure itself), and a term relative to the strength (of the
development of a project. Although this methodological rock matrix or of the rock mass) that can be mobilised.
approach is relatively limited at a national level, the Italian Such an approach is commonly used for the prevision of
LGP has surely spurred analogous reflections in other both “squeezing” behaviour (see for example, as reported
countries and one of the ITA Working Groups is also by Barla, 1998, the “Competence Indexes” proposed
specifically dedicated to the setting up of an international by Jethwa et al., 1984; Aydan et al., 1991; Singh et al.,
reference for tunnel design. 1992; Hoek and Marinos, 2000) and of rockburst
As remarked in the LGP, as well as in a previous paper phenomena (“Damage Indexes”: Russenes, 1974; Hoek and
(Russo et al., 1998), one of the elements of greatest Brown, 1980; Grimstad and Barton, 1993; etc.).
importance in design development is a clear distinction
between the concepts of geomechanical classes (or groups),
behaviour categories and technical classes of excavation.
In the following, some new considerations on the general
classification of the excavation behaviour are presented.

2 GENERAL CLASSIFICATION OF THE

EXCAVATION BEHAVIOUR

According to LGP, for the correct typological choice and

dimensioning of stabilisation measures, it is first of all
important to analyse the excavation behaviour of rock
masses, classified in homogeneous geomechanical groups,
under the existing stress conditions at tunnel level, in the
theoretical hypotheses of absence of any design
interventions. Fig.2.1: Example of rockburst (left: Loetschberg Tunnel,
Moreover, in general terms, a sufficiently complete Diederichs, 2005) and squeezing behaviour (right: deformation in
classification of the response of excavation requires the the pilot tunnel of the Fleres railway tunnel excavated by TBM
joint development of stress and geostructural analyses.
In practice, it is generally useful to make use of a
reasonably simplified approach which can offer a first A design alternative, which is relatively more accurate
general picture of the expected behaviour, after which it is than the index method, uses an analytical approach, for
necessary to carry out detailed verifications with more example the "convergence-confinement" method, to directly
rigorous and precise methods. From this point of view, quantify the response of the excavation in deformational
stress analyses are often performed to quantify some terms. For example, in the previously mentioned article by
the authors, a classification of the behaviour was presented
 Proceedings of the 11th Congress of International Society of Rock Mechanics, Lisbon, 9-13 July 2007, pp 979-982
in which 6 categories (a→f) were identified in function of
both the radial deformations at the excavation face and the
development of the plastic zone around the cavity (for
relative details, see the left part of Fig. 2.3).
In addition, by means of geostructural type analysis, it is
possible to focus better on the gravitational instabilities
connected to the real discontinuous character of the rock
masses.
On the other hand, the stress analyses on their own cannot
supply univocal indications of the expected deformation
phenomenon: for example, it is quite intuitive that a high
stress to strength ratio can differently determine
“squeezing” behaviour or "rockburst" according to real
geostructural conditions (Fig.2.1). For these reasons, as
previously mentioned, similar indexes (either of
competence or damage) are used for the empirical
forecasting of both phenomena.
Consequently, as a first classification, it is useful to refer
jointly to adequate geostructural indexes, especially when
they can be related to the consequent self-supporting
capacity of the rock masses (for example, the well-known
systems RMR of Bieniawski and the Q of Barton).
Accordingly, Fig.2.2 shows the basic scheme for a
general picture of the possible behaviour during excavation
in terms of typical deformation phenomenon.
Fig.2.2: Conceptual scheme for a general setting of the ground
behaviour upon excavation. The numbers in the box refer to the
attempt of association of stabilisation measures later described in
section 2.2. (..),/=eventual, alternative measure
2.1 Proposed classification system
Following the conceptual scheme of Fig.2.2, the
classification system currently adopted by the authors
combines the above mentioned behaviour categories with
the Bieniawski RMR classes (Fig.2.3, Russo et Grasso,
2006). As can be seen, the matrix that results from such a
double classification approach allows an optimal
focalisation of the specific design problem. Furthermore, a
rational choice of the type of stabilisation measures may be
derived as a function of the most probable potential
deformation phenomenon that is associated to the different
stress and geostructural combination.
It is, however, necessary to add that the proposed
classification matrix is not able, as it is logical, to cover all
the possible design criticalities and some particular
geological conditions have to be analysed separately (for
example, the presence of swelling material, complex
geostructural situations, etc.).
It is important to note that the proposed scheme may
supply also a further indication of the most suitable design
analysis, with assimilation either to a continuous or
discontinuous model as a function of the geostructural
characteristics of the rock masses.
The analysis for the forecasting of the excavation
behaviour may be developed in probabilistic terms,
incorporating the variability and uncertainty of the
geomechanical and boundary conditions. A practical
example is presented below.
The case refers to a stretch of a circular tunnel with a 5m
radius and 300m of overburden, realised in calcareous-
dolomitic formations. The input geomechanical data, mainly
based on the results of some boreholes, are summarised in
Table 2.1. The probabilistic distribution of the GSI
(Geological Strength Index, Hoek et al., 1995) represents
the best-fitting of the results of MonteCarlo simulation
graphically reported in Fig. 2.4. In particular, the
"quantitative" approach proposed by Cai et al. (2004), based
on the estimation of the block volume (Vb) and the Joint
Condition Factor (Jc) has been in this case applied. It should
be noted that these last input parameters have been
measured and statistically treated excluding fault zones, in
order to represent the "ordinary" rock mass condition.
Input
Distribution min max
GSI LGN(50,7) 35 67
σc (MPa) TRG(30,50,70) 30 70
mi UNF(8,12) 8 12
RMR BETA(12,12,84) 22 65
Output
RMR→ II III IV
CC↓
c 0.8% 58% 40%
d 0.4% 0.8%
Table 2.1 Main input parameters and output of the probabilistic
analysis of the excavation behaviour. Notes: LGN=lognormal;
TRG=triangular; UNF=uniform; σc=uniaxial compression strength
of the intact rock; mi=Hoek-Brown constant of the intact rock
(1980); CC=behavioural category (stress analysis).
The behavioural analyses were performed with the
“Convergence-Confinement” method (Carranza T.
solutions, 2004). The GSI is used to derive the rock mass
parameters from those of the intact rock by using for shear
strength and deformability, respectively the Hoek et al.
(2002) and the Hoek and Diederichs (2006) equations. From
the results of the probabilistic analysis, reported in Fig. 2.5
and Table 2.1, the following comments may be added:
● according to the stress analysis, the response to the
excavation is always elasto-plastic and in 99% of the
cases the intensity of the deformation phenomena is
relatively contained (“c” behaviour category); the
excavation face is consequently stable and the total radial
displacements, in the absence of stabilisation measures,
are limited to few centimetres;
● in almost all cases, the RMR class falls between the III
(58%) and the IV (41%): the geomechanical quality of
the rock mass is consequently fair to poor; taking into
account the diameter of the excavation, the relative self-
supporting rock mass properties are such as to require
increasingly more important measure of confinement
and/or reinforcement;
● the most important deformation phenomenon for
dimensioning the stabilisation measures is therefore that
 Proceedings of the 11th Congress of International Society of Rock Mechanics, Lisbon, 9-13 July 2007, pp 979-982

of a gravitational type, passing from the potential
detachment of rock wedges to purely caving behaviour in
the most unfavourable contexts. It is important to note
that, in function of the orientation of the discontinuities,
the gravitational collapses could also involve the
excavation face.
The detailed design, as a result, should include the use of
the most suitable methods for the analyses of a
discontinuous medium, applying, for example, solutions
based on the limit equilibrium theory and/or, preferably
numerical methods like distinct element (D.E.M.).

Fig.2.3: General classification scheme of the excavation behaviour

Notes: δo=radial deform. at the face; Rp/Ro=plastic radius/radius
of cavity; σθ=max tang. stress; σcm=rock mass strength. (i) The
limits of shadow zones are just indicative; (ii) in the brittle failure
domain (→rockburst) the deformation index of the stress analysis
can be intended just as indicators of the increasing potentiality of
the deformational phenomenon; (iii) according to Diederichs
(2005) the potentiality of the rockburst becomes relevant for rock
masses with Brittle Index IF=(σc/σt)>8 e σc>80MPa. In other
cases, the shear failure appears to be more probable and squeezing
is still the most typical deformation phenomenon.

Figure 2.5: Example of behaviour classification obtained through

probabilistic analysis.

2.2 Definition of the types of stabilisation measures

Figure 2.4: Example of probabilistic estimate (n.500 simulations)
of the GSI derived from the quantitative approach proposed by Cai
et al. (2004). Notes: M=Massive; (V)B=(Very) Blocky;
D=Disturbed; DS=Disintegrated; F/L/S=Foliated/ Laminated/
Sheared. (V)G=(Very)Good; F=Fair; (V)P=(Very)Poor.
The design choice of the stabilisation measures
(confinement, reinforcement, etc.) and therefore the
composition of the Section Type is a consequence of the
behavioural classification.
An example, in this sense, basically connected to
excavation by traditional techniques, is presented below.
In Table 2.3, the main design actions that can be applied
in the underground excavation have been schematically
summarised with some examples of the consequent
stabilisation measures. Furthermore, as above anticipated,
an attempt of association of these main design actions to
the different deformation phenomena is schematised in Fig.
2.2.
# Design action * Stabilization measure (example)
 Proceedings of the 11th Congress of International Society of Rock Mechanics, Lisbon, 9-13 July 2007, pp 979-982

1 In advancement to the excavation 3 CONCLUSION

1.1 Pre-confinament of Forepoling
instable wedges A framework for comprehending the expected excavation
1.2 Pre-reinforcement Pre-riqualification of rock mass behaviour is proposed, based on the combination of two
of rock mass contour by fully connected elements classification systems, in order to take into account both the
1.3 Pre-confinement of Sub-horizontal jet-grouting results of stress analysis, as well as the geostructural
the excavation contour canopy
characteristics of the rock mass and then its self-supporting
1.4 Tunnel face pre- Injected fiber-glass elements,
reinforcement jet-grouting, etc.
capacity. The matrix that results from the double
1.5 Pre-reinforcement Umbrella arch classification approach allows for first an optimal
of excavation contour focalisation of the design problem and then a rational choice
.. .. .. of the stabilisation measures in function of the most
2 During excavation probable potential deformation phenomenon.
2.1 Over-excavation to
allow convergences Acknowledgements: the authors would like to thank
2.2 Radial confinement Bolting (instable wedges Prof. Eng. Sebastiano Pelizza for his critical review of the
confinement); shotcrete (fiber- paper and for his numerous suggestions.
reinforced or with wire mesh);
steel ribs; etc. 4 BIBLIOGRAPHY
2.3 Rock mass Riqualification by means of
reinforcement fully connected elements Barla G. (1994): Metodi di analisi progettuale per gallerie in
2.4 De-confinement (to Sliding steel-ribs; joints and/or rocce spingenti. Atti del Quinto Ciclo di Conferenze di
allow convergence for deformable elements in the meccanica e ingegneria delle rocce pp 7.1-7.10
stress unloading) shotcrete Barton N., Lien R. and Lunde J. (1974). Engineering classification
2.5 Protection Anchored double-torsion steel of rock masses for the design of tunnel support. Rock
mesh Mechanics, vol.6, n.4
Table 2.3 Main design actions in underground excavation and Bieniawski Z.T. (1989): Engineering Rock Mass Classification,
examples of the typical associated stabilisation measures. John Wiley & Son.
Notes: *The drainage of water in advancement should be generally Cai M., Kaiser P.K., Uno H., Tasaka Y. and Minami M. (2004):
added when necessary. Estimation of Rock Mass Deformation Modulus and Strength of
Jointed Hard Rock Masses using the GSI system. International
As can be seen in Fig. 2.2, the design solutions are not Journal of Rock Mechanics and Mining Sciences n.41, pp.3-19.
always univocal and different stabilisation measures can Carranza-Torres C. (2004): Elasto-plastic solution of tunnel
theoretically be applied, measures whose application problems using the generalized form of the Hoek–Brown failure
possibilities should be opportunely verified with design criterion. Proceedings of the ISRM SINOROCK 2004.
calculations and then compared in terms of construction Diederichs M. (2005): General Report: Summary of Meetings with
time and costs. Geodata with recommendations towards a Design Methodology
A typical example concerns the stabilisation measures for spalling Failure and Rockburst Hazards. Personal
that can be chosen in the case of “squeezing” behaviour, communication to Geodata.
where may be necessary to evaluate (and compare) Hoek E. and Brown E.T. (1980). Underground Excavations in
contrasting procedures: Rock. The Institution of Mining and Metallurgy, London, 527 p.
(i) to allow a certain decompression of the rock mass; Hoek E. and Diederichs M. (2006): Estimation of rock mass
(ii) on the contrary, to prevent it with pre-reinforcement for modulus. Int.Journal of Rock Mechanics and Mining Science.
increasing the geomechanical properties of the rock mass at Hoek E. and Marinos P., (2000): Predicting Squeeze. Tunnels and
the excavation face and/or at the boundary, or Tunneling International, November, pp.45-51.
(iii) to apply some "hybrid" stabilisation solutions. Hoek E., Carranza-Torres C. and Corkum B. (2002): Hoek-Brown
The design calculations will allow to differentiate the failure criterion – 2002 Edition. Proc.North American Rock
intensity of the interventions, as a function of the detailed Mechanics Society. Toronto, July 2002.
analysis relative to the different geomechanical conditions Hoek E., Kaiser P.K. and Bawden W.F. (1995): Support of
classified in Fig. 2.3. Underground Excavations in Hard Rock. Balkema, Rotterdam,
As an example, the results of the probabilistic analysis 215pp.
presented in Table 2.1 would lead to the typological “NPSUW” National Project for Design, Tendering and
indications shown in Table 2.4. Construction Standards in Underground Works (promoted by
AGI, GEAM, IAEG, ITCOLD, SIG, SIGI), (1997): “Guidelines
Classification → c/III (≈ 60%) c/IV (≈40%) for design, tendering and construction of underground works” -
Section Type C1 C2 Attachment of Gallerie e Grandi Opere Sotterranee, No.51.
Stabilization -Eventual* bolting -Eventual** pre- Russo G., Kalamaras G.S. and Grasso P. (1998): A discussion on
measures in of tunnel face reinforcement of the concepts of geomechanical classes, behavior categories and
advancement tunnel face and technical classes for an underground project. Gallerie e grandi
crown opere sotterranee, N.54, pp.40-51.
Radial stabilisation -Sistematic bolting; -Steel ribs; Russo G. and Grasso P. (2006): Un aggiornamento sul tema della
measures -Shotcrete (fbr) -Shotcrete (fbr) classificazione geomeccanica e della previsione del
Table 2.4: Practical example of the typological choice of the comportamento allo scavo. Gallerie e grandi opere sotterranee,
stabilisation measures with reference to the case presented in Table N.80, pp.56-65.
2.1. Notes: fbr = fiber-reinforced; *in function of the possible
kinematic instability of rock wedges; **in function of the possible
kinematic instability and/or when the intervention allows an
optimisation of the construction times and costs.